Phase Shifts

A scattering experiment hurls a beam at a target and reads an angular pattern. Predicting the pattern seems to demand everything about the potential — depth, shape, wiggles. The astonishing bookkeeping fact of this chapter: for a central potential the wave leaving the collision differs from the free wave in exactly one real number per angular-momentum channel, the phase shift $\delta_\ell(k)$ . An elastic channel can neither gain nor lose probability — the potential can only slide the asymptotic wave train. Attraction pulls it in and advances the phase; repulsion pushes it out. Everything a detector registers is rebuilt from the list $\delta_0,\delta_1,\delta_2,\dots$ This article is where the chapter's matching machinery lives. We derive the master formula converting one interior number — the logarithmic derivative at the potential's edge — into $\tan\delta_\ell$ , repairing a booby trap in the base text along the way: the formula is true only for the log-derivative of the full radial function $R=u/r$ , not of $u$ , and the two differ by a real, computable error. We derive the hard-sphere phases and the square well's $\delta_0$ in closed form, prove the sign theorem (attraction ⇒ $\delta_\ell>0$ ) from an exact integral identity, and let the machine exhibit Levinson's staircase: each new bound state a deepening well swallows lifts the threshold phase by exactly $\pi$ . Cross sections built from the phases are quoted from Partial-Wave Expansion; the resonance sweep and the $k\to0$ story are owned by Resonances and the Breit-Wigner Formula and Low-Energy Scattering and the Scattering Length.

Key takeaways

Elastic scattering by a central potential reduces, channel by channel, to one real number: $u_\ell \to \sin(kr-\ell\pi/2+\delta_\ell)$ .
THE master matching formula: $\tan\delta_\ell=\dfrac{k\,j_\ell'(ka)-\gamma_\ell\,j_\ell(ka)}{k\,n_\ell'(ka)-\gamma_\ell\,n_\ell(ka)}$ , $\gamma_\ell=R'/R$ at the edge — the FULL radial function's log-derivative. Using $u'/u$ (the base text's slip) is wrong by exactly $1/a$ : machine-proven, repaired below.
Hard sphere: $\tan\delta_\ell=j_\ell(ka)/n_\ell(ka)$ , so $\delta_0=-ka$ exactly. Square well: $\delta_0=-ka+\arctan[(k/K)\tan Ka]$ — the bound-state matching of Particle in a Spherical Well and Infinite Spherical Box inherited at $E>0$ .
Sign theorem (exact): $\sin\delta_\ell=-\tfrac{2m}{\hbar^2}\int_0^\infty j_\ell(kr)\,V(r)\,u_\ell(r)\,r\,\mathrm{d}r$ for unit-amplitude $u_\ell$ — attraction pulls $\delta_\ell$ positive, repulsion negative.
Levinson's staircase (machine exhibit; theorem stated with source): $\delta_0(0)=n_b\pi$ on the continuous branch — our depth ladder returns $0,\pi,\pi,2\pi,3\pi$ for wells holding $0,1,1,2,3$ s-states.
KEY IDEA: the entire elastic effect of a central potential on channel $\ell$ is one real number $\delta_\ell(k)$ , manufactured at the edge by a single logarithmic derivative — and the phase list carries the potential's whole scattering identity, bound states included.

The core insight

In each angular-momentum channel the potential's entire effect on elastic scattering is one real number, the phase shift $\delta_\ell$ — the shift of the asymptotic radial wave relative to the free solution — and it is manufactured at the potential's edge by a single logarithmic derivative.

Where this lives today

Phase shifts are the negotiating table between scattering data and theory. Nucleon–nucleon partial-wave analyses still publish $\delta_\ell(E)$ tables, and every modern nuclear force — chiral effective field theory included — is graded on reproducing them. In hadron spectroscopy, resonance claims live or die by whether a partial wave's phase actually rises through $\pi/2$ (formalized on the Argand circle of The S-Matrix and Unitarity in Scattering). In ultracold-atom laboratories the whole interatomic interaction is one s-wave phase, tuned through Feshbach resonances — $\delta_0$ as a control knob. Whenever the interaction is unknown but rotationally symmetric, the honest interface is the phase list.

The physical picture

Watch one channel at a time. The free $\ell=0$ radial function $u_0=\sin kr$ is a wave train tied to the origin. Switch on an attractive well of range $a$ : inside, the local wavenumber grows from $k$ to $K=\sqrt{k^2+2mV_0/\hbar^2}$ , so the wave oscillates faster where the potential holds it and re-emerges carrying extra phase. The exterior — still a free sinusoid — arrives "early": every node sits closer to the origin than its free counterpart by the same distance $\delta_0/k$ . That uniform slide is the phase shift, and its sign remembers the potential's: attraction pulls the pattern in ( $\delta_0>0$ ), repulsion pushes it out.

Figure

Pull-in, push-out: the phase shift as a slide of the asymptotic wave train

Why one number and not a whole function? Probability is conserved channel by channel: the exterior is a superposition of incoming and outgoing spherical waves, and — as Spherical Hankel Functions and Outgoing Waves proves with its flux certificate — elastic flux balance forces their magnitudes equal. The only freedom left is the relative phase of the outgoing wave, and that is $2\delta_\ell$ . (Organizing the phases into an operator is The S-Matrix and Unitarity in Scattering's business.) The second organizing fact is reach: the centrifugal barrier holds high- $\ell$ waves away from a short-range potential, so at low energy only the smallest $\ell$ acquire any phase. And because $\delta_\ell$ is read off in the free region it is blind to interior detail: two wells with the same edge log-derivative scatter identically — one number to compute, and a warning that data cannot reconstruct the interior. The most surprising cargo the phases carry is the bound-state count:

Figure

Levinson staircase: the threshold phase counts bound states in exact units of pi

One number per channel

Far outside the potential the radial equation is free — Asymptotic Form of the Scattering Wave Function established what free means at large $r$ — and the potential's entire elastic legacy is a slide of that sinusoid. This is the article's defining equation; everything else is machinery for computing the one number in it.

u_\ell(r)\;\xrightarrow{\;r\to\infty\;}\;\sin\!\Big(kr-\frac{\ell\pi}{2}+\delta_\ell\Big) \qquad (1)

where: $u_\ell=rR_\ell$ , $k=\sqrt{2mE}/\hbar$ , the $-\ell\pi/2$ is the free centrifugal phase (present even at $V=0$ , per Spherical Hankel Functions and Outgoing Waves), and $\delta_\ell$ is the extra phase the potential manufactures. Equivalently: the outgoing half of the wave is multiplied by $S_\ell=e^{2i\delta_\ell}$ — the pure-phase response whose operator theory is owned by The S-Matrix and Unitarity in Scattering. Once the phases are known, the observables follow by quotation. From the Rayleigh decomposition of the beam, Partial-Wave Expansion derives the amplitude ladder and the cross-section sum

f(\theta)=\frac{1}{k}\sum_{\ell=0}^{\infty}(2\ell+1)\,e^{i\delta_\ell}\sin\delta_\ell\,P_\ell(\cos\theta) \qquad (2)

\sigma=\frac{4\pi}{k^2}\sum_{\ell=0}^{\infty}(2\ell+1)\,\sin^2\delta_\ell \qquad (3)

where: both formulas are that article's property — quoted, not re-derived. They define this article's job: produce the $\delta_\ell(k)$ that feed them. Division of labor: Partial-Wave Expansion decomposes, this article computes the phases, The S-Matrix and Unitarity in Scattering constrains them, Resonances and the Breit-Wigner Formula reads their sweeps, Low-Energy Scattering and the Scattering Length takes $k\to0$ .

Reading the potential at its edge

For a potential negligible beyond radius $a$ , computing $\delta_\ell$ collapses to a boundary ritual: solve the interior however the potential demands; the exterior learns one number, the log-derivative at the edge. This machinery — with the hard-sphere formula — was deeded here by Spherical Hankel Functions and Outgoing Waves ("the base article's matching-to-phase-shift and pole derivations are theirs by chapter ownership; relocation noted in the provenance"): the matching derivation lands here, the pole derivation in Bound States and Zero-Energy Resonances.

Derivation AThe master matching formula at the potential edge

We derive the formula converting one interior number into the phase shift, all channels at once. The strategy is 1D matching transplanted to the radial problem: write the free exterior as a phase-shifted combination of the two spherical solutions, demand continuity of value and derivative at $r=a$ , and watch only the logarithmic derivative survive the ratio. The one genuine subtlety — the base text's trap — is which log-derivative: the formula is married to the full radial function $R=u/r$ , and feeding it $u'/u$ produces phase shifts that are simply wrong.

Step 1. Write the exterior in the phase-shifted basis. For $r>a$ the radial equation is free, and the combination producing exactly the asymptotic form (1) is

R_\ell(r)=\cos\delta_\ell\; j_\ell(kr)-\sin\delta_\ell\; n_\ell(kr), \qquad r>a

Why this combination: asymptotically $j_\ell\to\sin(kr-\ell\pi/2)/kr$ and $n_\ell\to-\cos(kr-\ell\pi/2)/kr$ , so the sum collapses to $\sin(kr-\ell\pi/2+\delta_\ell)/kr$ . At $\delta_\ell=0$ the irregular piece vanishes — the pure- $j_\ell$ free solution of The Free Particle in Three Dimensions; with a potential occupying $r<a$ the origin is outside the free domain and $n_\ell$ is legitimate.

Step 2. Match the logarithmic derivative. The interior solution (regular at the origin) arrives at the edge carrying $\gamma_\ell\equiv R'_\ell(a^-)/R_\ell(a^-)$ , and continuity of $R$ and $R'$ forces the exterior to match it:

\gamma_\ell=\frac{k\big[\cos\delta_\ell\, j_\ell'(ka)-\sin\delta_\ell\, n_\ell'(ka)\big]}{\cos\delta_\ell\, j_\ell(ka)-\sin\delta_\ell\, n_\ell(ka)}

Why the log-derivative: normalization is physically empty, so the edge transmits exactly one meaningful number — the whole interior compressed into $\gamma_\ell$ .

Step 3. Solve for the phase. Cross-multiplying and collecting the $\cos\delta_\ell$ and $\sin\delta_\ell$ terms:

\tan\delta_\ell=\frac{k\,j_\ell'(ka)-\gamma_\ell\, j_\ell(ka)}{k\,n_\ell'(ka)-\gamma_\ell\, n_\ell(ka)} \qquad (4)

This is THE master formula. Hard sphere: $R_\ell(a)=0$ means $\gamma_\ell\to\infty$ , and (4) collapses to

\tan\delta_\ell=\frac{j_\ell(ka)}{n_\ell(ka)} \qquad (5)

For $\ell=0$ , $j_0=\sin x/x$ and $n_0=-\cos x/x$ give $\tan\delta_0=-\tan ka$ : the hard sphere's s-wave phase is $\delta_0=-ka$ , exactly, at every energy — the excluded region deletes $a$ worth of wave train. No interaction: a free interior arrives with the log-derivative of $j_\ell$ itself, the numerator vanishes identically, $\delta_\ell=0$ .

Step 4. The convention ledger — because the chapter's raw source gets this wrong. Formula (4) is written in the $R=u/r$ dialect. If you carry $\beta_\ell=u'(a)/u(a)$ instead (natural when you solved the 1D-looking $u$ -equation), the two differ by the geometric term from $u=rR$ : $\beta_\ell=\gamma_\ell+1/a$ . Either convert, or use the Riccati dialect built for $u$ — with $\hat s_\ell(x)=x\,j_\ell(x)$ , $\hat c_\ell(x)=-x\,n_\ell(x)$ :

\tan\delta_\ell=\frac{\beta_\ell\,\hat s_\ell(ka)-k\,\hat s_\ell'(ka)}{k\,\hat c_\ell'(ka)-\beta_\ell\,\hat c_\ell(ka)}

Why it matters: for a square well at $k=0.6$ , $a=1$ , $2mV_0a^2/\hbar^2=6$ , feeding $\beta_\ell=u'/u$ into (4) returns $\delta_0=-0.832$ rad where the truth is $-0.768$ — an 8% phase error. Only at $\gamma_\ell\to\infty$ (hard wall) do the dialects agree, which is exactly why the slip survives casual checks. *Verified: (4) against direct radial-ODE integration for the hard sphere, square well (three energies) and Gaussian well (two matching radii, $10^{-7}$ ); the $\gamma\to\infty$ collapse to (5); the $u'/u$ trap and its $\beta=\gamma+1/a$ repair machine-proven, incl. the sympy identity that the dialect formulas coincide under $\beta a=\gamma a+1$ (article verifier, 46 checks).*

The error is plausible: the $u$ -equation looks exactly like 1D Schrödinger, so $u'/u$ feels like the natural boundary datum — and it is, provided you match against Riccati–Bessel exteriors rather than bare $j_\ell,n_\ell$ . The base text printed (4) with $\beta_\ell=u'/u$ in the $\gamma_\ell$ slot; the verifier proves that combination wrong for every finite log-derivative. The cheat sheet carries the trap line.

The attractive square well, inherited at positive energy

Particle in a Spherical Well and Infinite Spherical Box solved the bound states from the s-wave matching $k\cot kR=-\kappa$ and closed with a deed: the " $E>0$ face of the same well — phase shifts, resonances, the zero-energy story — belongs BY NAME to the scattering chapter…, which inherits this article's matching setup at $E>0$ ." We cash the inheritance.

Go deeper · Derivation B · 1 minSquare-well s-wave phase from the inherited matching

Derivation BSquare-well s-wave phase from the inherited matching

We compute $\delta_0(k)$ in closed form for the attractive well $V=-V_0$ ( $r<a$ ) — the canonical case where the master formula produces an explicit, exactly checkable function. The route reuses the bound-state problem's interior, a regular sine of shifted wavenumber, matched now to a phase-shifted exterior instead of a decaying tail. Watch the answer split into a geometric piece ( $-ka$ ) and a dynamical arctangent, and watch the parent's bound-state condition reappear at the other energy sign.

Step 1. Interior. Regularity at the origin forces $u_{\rm in}=A\sin Kr$ with $K=\sqrt{k^2+2mV_0/\hbar^2}$ ; its edge datum is $\beta_0=u'/u\big|_a=K\cot Ka$ .

Step 2. Exterior and matching. For $\ell=0$ the Riccati dialect is elementary ( $\hat s_0=\sin x$ , $\hat c_0=\cos x$ ): the exterior is $u_{\rm out}=C\sin(kr+\delta_0)$ , and matching $u'/u$ gives

k\cot(ka+\delta_0)=K\cot Ka

Why this is the parent's equation: at $E<0$ the exterior sine becomes a decaying exponential and the same matching reads $k\cot kR=-\kappa$ — one analytic family, two energy signs.

Step 3. Solve.

\delta_0(k)=-ka+\arctan\!\Big[\frac{k}{K}\tan Ka\Big] \qquad (6)

where: the $-ka$ is the hard-sphere-like geometric slide (the exterior must reorganize across the region the well occupies) and the arctangent carries the dynamics — for a weak well it exceeds $ka$ slightly, leaving $\delta_0$ small and positive: attraction pulls the wave in. Why sign statements need branch language: at $k=0.8$ , $2mV_0a^2/\hbar^2=6$ the principal-value arctangent gives $-0.99$ rad; the continuous-branch value is $+2.15$ — same angle mod $\pi$ . One bound state ⇒ the branch anchored at $\delta_0(\infty)=0$ starts at $\pi$ . *Verified: (6) against ODE integration at $k=0.3,0.8,1.5$ ; the $\ell=1$ analogue via interior $j_1(Kr)$ matching; the continuous-branch $+2.147$ at $k=0.8$ (verifier).*

The sign theorem, exactly

Go deeper · Derivation C · 1 minAttraction pulls the phase forward: the exact integral identity

Derivation CAttraction pulls the phase forward: the exact integral identity

We prove $\mathrm{sign}(\delta_\ell)=-\,\mathrm{sign}(V)$ for definite-sign potentials via an exact comparison of the true radial wave against the free one — the Wronskian trick behind every comparison theorem in this subject. The result doubles as the exact parent of the Born phase shift, whose systematics live in their own article.

Step 1. The two equations. The free Riccati function $\hat s_\ell(kr)=kr\,j_\ell(kr)$ and the true reduced wave $u_\ell$ (normalized to unit asymptotic amplitude, $u_\ell\to\sin(kr-\ell\pi/2+\delta_\ell)$ ) satisfy

\hat s_\ell''+\Big(k^2-\frac{\ell(\ell+1)}{r^2}\Big)\hat s_\ell=0, \qquad u_\ell''+\Big(k^2-\frac{2mV}{\hbar^2}-\frac{\ell(\ell+1)}{r^2}\Big)u_\ell=0

Step 2. Cross-multiply and integrate. Multiplying the first by $u_\ell$ , the second by $\hat s_\ell$ , and subtracting kills everything but the potential term:

\frac{\mathrm{d}}{\mathrm{d}r}\big(\hat s_\ell u_\ell'-\hat s_\ell' u_\ell\big)=\frac{2mV}{\hbar^2}\,\hat s_\ell\, u_\ell

Why the boundary terms cooperate: at the origin both functions vanish as $r^{\ell+1}$ ; at infinity the Wronskian of two unit sinusoids differing by $\delta_\ell$ evaluates to $-k\sin\delta_\ell$ by angle subtraction.

Step 3. Read off the identity.

\sin\delta_\ell=-\frac{2m}{\hbar^2}\int_0^\infty j_\ell(kr)\,V(r)\,u_\ell(r)\,r\,\mathrm{d}r \qquad (7)

where: this is exact — no weak-potential assumption — for $u_\ell$ at unit asymptotic amplitude. A normalization booby trap rides with it: the chapter's raw source prints the prefactor as $2mk/\hbar^2$ , which is only consistent if $u_\ell$ carries amplitude $1/k$ ; with unit amplitude the extra $k$ is provably wrong — the verifier checks the identity at two different energies, and a coefficient with a stray $k$ cannot match both.

Step 4. Extract the sign, then the Born limit. Where a short-range potential lives, $j_\ell\,u_\ell\,r$ is positive (both functions rise from the origin as $r^{\ell+1}$ , no node yet), so the integral takes the sign of $V$ : attraction makes $\sin\delta_\ell>0$ , repulsion negative. Replacing $u_\ell\to kr\,j_\ell(kr)$ — legitimate when the wave is barely distorted — gives $\delta_\ell\approx-(2mk/\hbar^2)\int Vj_\ell^2r^2\mathrm{d}r$ , the first-order phase owned by Born Approximation. Verified: identity (7) for attractive and repulsive Gaussian wells, $\ell=0,1$ (four sign checks, integral matching $\sin\delta$ to $2\times10^{-3}$ ), and at a second energy — the coefficient is $k$ -independent, killing the stray- $k$ variant (verifier).

Counting bound states without solving for them

Sweep a well's depth and watch the threshold phase. The machine ran the ladder $2mV_0a^2/\hbar^2=1,6,18,44,80$ : the continuous-branch $\delta_0(k\to0)$ came out $0,\pi,\pi,2\pi,3\pi$ — exactly $\pi$ per bound state, including the refusal to advance between the two depths holding a single s-state. This is Levinson's theorem (N. Levinson, 1949, Kgl. Danske Vid. Selsk. Mat.-fys. Medd. 25, no. 9):

\delta_\ell(0)-\delta_\ell(\infty)=n_\ell\,\pi \qquad (8)

where: $n_\ell$ counts bound states of angular momentum $\ell$ ; $\delta_\ell(\infty)=0$ pins the branch. The s-wave carries famous fine print — a half-bound state exactly at threshold contributes $\pi/2$ — and that creature, with the proof of (8) by pole-counting in the complex $k$ -plane, belongs to Bound States and Zero-Energy Resonances, which owns the analytic continuation this staircase secretly performs. We state the theorem with its source and exhibit it by machine; we do not prove it. Two corollaries: scattering data counts bound states, and the theorem disciplines every fit — a $\delta_0(E)$ drifting to the wrong multiple of $\pi$ asserts the wrong spectrum however well it matches $\sigma$ . And as $k\to0$ the matching formula forces $\tan\delta_\ell\propto(ka)^{2\ell+1}$ , shutting every channel but the s-wave (exponents machine-checked; the coefficient — the scattering length — is owned by Low-Energy Scattering and the Scattering Length).

Limiting cases

case	result	meaning
hard wall, $\gamma_\ell\to\infty$	$\tan\delta_\ell=j_\ell(ka)/n_\ell(ka)$ ; $\delta_0=-ka$	excluded region deletes wave train
no potential	$\delta_\ell=0$ identically	master formula's numerator vanishes
weak potential	$\delta_\ell\approx-\frac{2mk}{\hbar^2}\int V j_\ell^2 r^2\mathrm{d}r$	first-order limit of identity (7); owner Born Approximation
$k\to0$	$\tan\delta_\ell\propto(ka)^{2\ell+1}$	centrifugal shutdown; coefficient owned by Low-Energy Scattering and the Scattering Length
$\delta_\ell$ sweeps $\pi/2$ fast	$\sin^2\delta_\ell$ spikes to 1	quasi-bound state; lineshape owned by Resonances and the Breit-Wigner Formula
well deepens through a capture	$\delta_0(0)$ jumps by $\pi$	Levinson staircase; pole mechanics owned by Bound States and Zero-Energy Resonances

Common trapA phase shift is one number mod pi — until physics forces you onto the continuous branch

The matching formula delivers only $\tan\delta_\ell$ , so a computer reports $\delta_\ell$ in $(-\pi/2,\pi/2]$ , and cross sections — built from $\sin^2\delta_\ell$ — are perfectly happy. Three statements are not: Levinson's theorem (the staircase lives on the branch anchored at $\delta_\ell(\infty)=0$ ), the sign theorem (our one-bound-state well has $\delta_0=+2.15$ , which an arctangent reports as $-0.99$ — "repulsive"!), and any resonance fit needing the phase to rise through $\pi/2$ . Unwrap first, interpret second. And $\delta_\ell$ is the phase beyond the free $-\ell\pi/2$ , not the total.

Worked examples

Example 1 — hard sphere at $ka=0.5$ : three phases and a cross section. A beam meets an impenetrable sphere with $ka=0.5$ — wavelength comparable to the obstacle.

Step 1. Phases from formula (5): $\delta_0=-ka=-0.5$ rad $=-28.65°$ exactly; $\tan\delta_1=j_1(0.5)/n_1(0.5)$ gives $\delta_1=-2.08°$ ; $\delta_2=-0.037°$ . The centrifugal gate is brutal — by $\ell=3$ the phase is below $5\times10^{-6}$ rad.

Step 2. Assemble the quoted sum (3): $\sigma=11.75\,a^2=3.74\,\pi a^2$ , the s-wave carrying $98.3\%$ , the p-wave $1.7\%$ . The value sits well above the geometric shadow $\pi a^2$ — at threshold it heads to $4\pi a^2$ , the wave-not-ray surprise whose derivation belongs to Low-Energy Scattering and the Scattering Length.

Answer: $\delta_0=-28.65°$ , $\delta_1=-2.08°$ , $\delta_2=-0.037°$ ; $\sigma=3.74\,\pi a^2$ , 98.3% s-wave.

Example 2 — a deuteron-class well meets a 1 MeV neutron. Model the np triplet interaction as a square well of range $R=2.0$ fm, depth $V_0=36$ MeV. With reduced mass $\mu=m_pm_n/(m_p+m_n)=469.46$ MeV $/c^2$ , the radial unit is $\hbar^2/2\mu=41.47$ MeV·fm².

Step 1. Credential the well: the parent article's bound-state matching gives one s-level at $B=2.03$ MeV — deuteron-class.

Step 2. Kinematics: cm energy is $E_{\rm lab}/2$ ; at $E_{\rm lab}=1$ MeV, $k=0.110$ fm⁻¹.

Step 3. Phases from (6) on the continuous branch ( $n_b=1$ ⇒ start at $\pi$ ): $\delta_0=147.1°/117.5°/102.0°$ at 1/5/10 MeV — a phase born at $180°$ and marching down, the Levinson-mandated descent for a one-bound-state channel.

Step 4. At 1 MeV, $\sigma_0=(4\pi/k^2)\sin^2\delta_0=308$ fm² $=3.08$ b for this triplet channel alone. The measured thermal np $\approx20$ b needs the singlet channel and spin statistics — bookkeeping owned by Low-Energy Scattering and the Scattering Length.

Answer: $B=2.03$ MeV; $\delta_0=147.1°/117.5°/102.0°$ at $E_{\rm lab}=1/5/10$ MeV; $\sigma_{\rm triplet}(1\,\mathrm{MeV})=3.08$ b.

Origins · 1 minPhase Shift and how this was discovered

Phase Shift

1927–1957 · Uppsala, Oslo, Bristol

The phase shift became physics in 1927, when Hilding Faxén and Johannes Holtsmark transplanted Rayleigh's 1872 acoustic partial-wave analysis into the new quantum mechanics to explain a scandal: argon was nearly transparent to slow electrons — the Ramsauer–Townsend effect — which no classical cross section could arrange. Their channel-by-channel phases made the transparency an interference fact. Nevill Mott and Harrie Massey's 1933 treatise The Theory of Atomic Collisions codified the method and made $\delta_\ell$ the working currency of collision physics. After the war the currency went hard: nucleon–nucleon experiments were digested into phase-shift analyses — the 1957 Stapp–Ypsilantis–Metropolis fits the template — turning $\delta_\ell(E)$ tables into the primary interface between scattering data and nuclear forces, a role they still play in hadron spectroscopy.

History

1872

— Rayleigh's Theory of Sound decomposes waves scattering off a sphere into spherical harmonics, the partial-wave template

1927

— Faxén and Holtsmark compute quantum phase shifts for electron–atom scattering, explaining the Ramsauer–Townsend transparency

1933

— Mott and Massey's Theory of Atomic Collisions codifies phase-shift methodology

1949

— Levinson proves the threshold theorem tying the zero-energy phase to the bound-state count

1957

— Stapp, Ypsilantis and Metropolis publish the template nucleon–nucleon phase-shift analysis

Going further

References. Griffiths & Schroeter, Introduction to Quantum Mechanics, 3rd ed., Ch. 11 · Sakurai & Napolitano, Modern Quantum Mechanics, 2nd ed., §6.4 · Shankar, Principles of Quantum Mechanics, 2nd ed., Ch. 19 · Cohen-Tannoudji, Diu & Laloë, Quantum Mechanics, Vol. II, Ch. VIII · Taylor, Scattering Theory, Ch. 11 · Newton, Scattering Theory of Waves and Particles, 2nd ed.

Provenance. All formulae are standard partial-wave results. Verifier should: (1) confirm $f(\theta)=\frac1k\sum(2l+1)e^{i\delta_l}\sin\delta_l P_l$ and $\sigma=\frac{4\pi}{k^2}\sum(2l+1)\sin^2\delta_l$ against Griffiths Ch.11, Sakurai §6.4, Shankar Ch.19; (2) reproduce $\delta_0=-ka$ and $\sigma\to4\pi a^2$ for the hard sphere; (3) check the square-well matching $\tan(ka+\delta_0)=(k/k')\tan(k'a)$ (Cohen-Tannoudji Vol.2 §VIII-C); (4) verify the Born phase $\delta_l\approx-\tfrac{2mk}{\hbar^2}\int V j_l^2 r^2\mathrm{d}r$ ; (5) the n-p scattering lengths $a_t\approx5.4$ fm, $a_s\approx-23.7$ fm are standard nuclear-physics values (Krane, Introductory Nuclear Physics). (Lane note: items (1)–(4) executed in the verify script; the base's Eq. (8) mixed log-derivative dialects and is repaired in Derivation A with a machine proof; its "exact relation" prefactor $2mk/\hbar^2$ hides a $1/k$ normalization, pinned in Derivation C; item (5)'s np numbers are quoted at their owner, Low-Energy Scattering and the Scattering Length.)

Builds on: Partial-Wave Expansion · Asymptotic Form of the Scattering Wave Function · Scattering Amplitude Leads to: The S-Matrix and Unitarity in Scattering · Optical Theorem · Resonances and the Breit-Wigner Formula · Differential and Total Cross Section

Is a phase shift actually measurable, or just a theorist's bookkeeping?

Measurable, but only jointly. A differential cross section determines $|f(\theta)|^2$ ; inverting Eq. (2) is the daily work of partial-wave analysis — measure at enough angles, expand in Legendre polynomials, solve for the active phases. Sign and branch ambiguities are broken by interference observables (Coulomb–nuclear interference is the classic reference). What is not measurable is the interior wavefunction: many interiors, one $\delta_\ell$ .

Why does one real number per channel exhaust the physics — where did the rest of the wavefunction go?

Elastic unitarity: probability conservation forces the incoming and outgoing magnitudes equal (the flux argument of Spherical Hankel Functions and Outgoing Waves, formalized in The S-Matrix and Unitarity in Scattering), and a unit-modulus outgoing coefficient is one real phase. Everything else is compressed into the edge log-derivative; the master formula is the funnel from $\gamma_\ell$ to that phase.

Do I use the log-derivative of u or of R in the matching formula?

Of $R=u/r$ — formula (4) is married to it. If your solver hands you $\beta=u'/u$ , convert with $\gamma=\beta-1/a$ or use the Riccati-dialect formula in Derivation A, Step 4. Feeding $\beta$ straight in is the base text's error: at $k=0.6$ , $a=1$ it shifts $\delta_0$ from $-0.768$ to $-0.832$ rad. The hard sphere forgives the slip (both log-derivatives infinite), which is why it goes unnoticed.

My code returned a negative phase for an attractive well. Is the sign theorem wrong?

Check the branch first. Arctangents report $\delta$ in $(-\pi/2,\pi/2]$ ; with $n_b$ bound states Levinson puts the true low-energy phase near $n_b\pi$ , so the mod- $\pi$ representative can land negative. Our worked well at $k=0.8$ : $-0.99$ rad raw, $+2.15$ on the continuous branch. Unwrap from high energy down, then apply the theorem.

Does any of this work for the Coulomb potential?

Not as stated: everything here assumed a free exterior, which needs $V$ dying faster than $1/r$ (Asymptotic Form of the Scattering Wave Function derives the ledger). Bare Coulomb distorts the phase logarithmically forever, so constant $\delta_\ell$ 's never form; one defines Coulomb-modified phases against Coulomb wavefunctions instead. Screening (every real plasma or atom) restores a finite range and lands back here — the screened story is Scattering by a Screened Coulomb Potential's.

What exactly happens to δ₀ at the moment a well first binds a state?

At the critical depth the well supports a zero-energy solution flat at infinity: the scattering length diverges, the threshold phase sits exactly $\pi/2$ above its old plateau (Levinson's half-bound fine print), and an infinitesimally deeper well converts the threshold creature into a bound state, completing the $\pi$ step. The mechanics — an S-matrix pole marching up the imaginary axis through $k=0$ — is Bound States and Zero-Energy Resonances' story; the observables are Low-Energy Scattering and the Scattering Length's.